CN113592194A - Establishment of CO2Method of throughput effect prediction model and CO2Method for evaluating throughput effect - Google Patents

Establishment of CO2Method of throughput effect prediction model and CO2Method for evaluating throughput effect Download PDF

Info

Publication number
CN113592194A
CN113592194A CN202110966381.9A CN202110966381A CN113592194A CN 113592194 A CN113592194 A CN 113592194A CN 202110966381 A CN202110966381 A CN 202110966381A CN 113592194 A CN113592194 A CN 113592194A
Authority
CN
China
Prior art keywords
throughput
neural network
model
training
data set
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202110966381.9A
Other languages
Chinese (zh)
Inventor
岳明
宋田茹
朱维耀
刘昀枫
宋洪庆
孔德彬
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
University of Science and Technology Beijing USTB
Original Assignee
University of Science and Technology Beijing USTB
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by University of Science and Technology Beijing USTB filed Critical University of Science and Technology Beijing USTB
Priority to CN202110966381.9A priority Critical patent/CN113592194A/en
Publication of CN113592194A publication Critical patent/CN113592194A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/06—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons
    • G06N3/061—Physical realisation, i.e. hardware implementation of neural networks, neurons or parts of neurons using biological neurons, e.g. biological neurons connected to an integrated circuit
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Data Mining & Analysis (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Health & Medical Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Molecular Biology (AREA)
  • Evolutionary Computation (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Business, Economics & Management (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Human Resources & Organizations (AREA)
  • Neurology (AREA)
  • Strategic Management (AREA)
  • Economics (AREA)
  • Evolutionary Biology (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Microelectronics & Electronic Packaging (AREA)
  • Game Theory and Decision Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The present disclosure provides a method of establishing CO2Method of throughput effect prediction model and CO2A method for evaluating throughput effect. Said establishing CO2The method of the throughput effect prediction model comprises the steps of firstly obtaining CO2Handling a developed sample data set, dividing the sample data set into a training set and a testing set, setting a neural network structure, initializing hyper-parameters of a neural network model, training the model, and controlling a loss function value of the model within an error range by modifying the hyper-parameters, thereby obtaining a training resultA good model. The CO is2The method for evaluating the throughput effect comprises the following steps: obtaining target well area CO2Inputting the influencing factor parameter data set of the target well area into CO2And a throughput effect prediction model for predicting the oil change rate. The method can be applied to treat CO under different influence factor parameters2Rapid prediction of oil change rate of wells developed throughout the year.

Description

Establishment of CO2Method of throughput effect prediction model and CO2Method for evaluating throughput effect
Technical Field
The present disclosure relates to the field of oilfield development technologies, and in particular, to a CO2Method for establishing throughput effect prediction model and CO2A method for evaluating throughput effect.
Background
The ultra-low permeability oil reservoir is one of the important objects of oil and gas exploration and development in China at present, has the characteristics of low permeability, poor oil reservoir physical property, low single-well productivity and the like, and usually adopts CO2And (4) mining in a gas drive mode such as flooding. For conventional water-flooding and other exploitation modes, a set of mature oil field development evaluation standards are formed at present, but CO is aimed at2The evaluation index of the oil displacement effect needs to be improved.
Disclosure of Invention
In one aspect, a method of establishing CO is provided2A method of throughput effectiveness prediction modeling, the method comprising:
obtaining CO2Taking the impact factor parameter data set and the target parameter data set developed in a throughput manner as sample data sets; the parameters of the influencing factors comprise the soaking time, the fracture interval, the water saturation, the formation pressure, the porosity, the permeability, the fracture length, the injection amount and the single-layer thickness, and the target parameters comprise the oil change rate;
dividing the sample data set into a training set and a test set;
building a neural network model, setting weights of neurons in the neural network model, and setting an activation function and an optimizer;
inputting the training set into the neural network model for training;
if the loss function value of the neural network model is larger than a preset error, modifying the training times, the activation function, the optimizer and the Dropout ratio, and training the model again; if the loss function value of the neural network model is within a preset error range, stopping training and taking the trained neural network model as CO2And (4) a throughput effect prediction model.
In at least one embodiment of the present disclosure, the establishing the CO is performed before the dividing the sample data set into a training set and a test set2The method of the throughput effectiveness prediction model further comprises: and processing the sample data set into a sample data set which can be used by machine learning, and performing normalization processing on data in the processed sample data set.
In at least one embodiment of the present disclosure, the dividing the sample data set into a training set and a test set includes: dividing a training set and a testing set by using a train _ test _ split function in a python open source library skleern; the first 80% of the sample data set is used as a training set, and the last 20% is used as a test set.
In at least one embodiment of the disclosure, the built neural network model comprises an input layer, five hidden layers and an output layer which are sequentially connected from an input end to an output end; wherein each hidden layer comprises 200 neurons.
In at least one embodiment of the present disclosure, the setting weights of neurons in the neural network model, selecting an activation function and an optimizer includes: and setting the weight of the neuron in the neural network model by adopting an Xavier method, adopting a ReLU function as an activation function, and selecting an Adam optimizer.
In at least one embodiment of the present disclosure, the loss function of the neural network model comprises a mean square error; if the mean square error value of the neural network model is greater than 10-2Modifying the training times, the activation function, the optimizer and the Dropout ratio, and training the model again; if said spiritMean square error value through network model is less than or equal to 10-2Stopping training and using the trained neural network model as CO2And (4) a throughput effect prediction model.
In at least one embodiment of the present disclosure, the establishing CO2The method of the throughput effectiveness prediction model further comprises:
inputting the test set into the CO2Predicting the oil change rate in the huff and puff effect prediction model;
if the mean square error value of the predicted oil change rate is less than 10-2And determines the coefficient R2A value greater than 90%, then the CO2The prediction capability of the throughput effect prediction model meets the prediction requirement.
In another aspect, a CO is provided2A throughput effect evaluation method, the method comprising:
obtaining target well area CO2The method comprises the steps of taking out and developing an influence factor parameter data set, wherein the influence factor parameters comprise soaking time, fracture spacing, water saturation, formation pressure, porosity, permeability, fracture length, injection amount and single-layer thickness;
inputting the influencing factor parameter dataset of the target well region into CO2A huff and puff effect prediction model for predicting the oil change rate; the CO is2The method for predicting the throughput effect is to establish CO according to any one of claims 1 to 72The model is established by a throughput effect prediction model method.
In at least one embodiment of the present disclosure, the CO2The throughput effect evaluation method further includes: and (3) analyzing the sensitivity degree of the oil change rate to different influencing factor parameters by using a Sobol sensitivity analysis method.
In yet another aspect, a CO is provided2Throughput performance assessment apparatus comprising a processor and a memory having stored therein computer program instructions adapted to be executed by the processor, the computer program instructions when executed by the processor performing the CO according to any of the embodiments described above2And a step in the throughput effect evaluation method.
Drawings
The accompanying drawings, which are included to provide a further understanding of the disclosure and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the disclosure and together with the description serve to explain the principles of the disclosure.
FIG. 1 is a diagram of CO establishment according to some embodiments2A flow chart of a method of a throughput effect prediction model;
FIG. 2 is another CO establishment in accordance with some embodiments2A flow chart of a method of a throughput effect prediction model;
FIG. 3 is a CO according to some embodiments2A flow chart of a throughput effect evaluation method;
FIG. 4 is another CO according to some embodiments2A flow chart of a throughput effect evaluation method;
FIG. 5 is a CO according to some embodiments2A first-order sensitivity analysis schematic diagram of the throughput effect evaluation method;
FIG. 6 is a CO according to some embodiments2A full-order sensitivity analysis schematic diagram of the throughput effect evaluation method;
FIG. 7 is a CO according to some embodiments2Schematic diagram of a throughput effect evaluation device.
Reference numerals:
100-CO2throughput effect evaluation device, 101-processor, 102-memory
Detailed Description
The present disclosure will be described in further detail with reference to the drawings and embodiments. It is to be understood that the specific embodiments described herein are for purposes of illustration only and are not to be construed as limitations of the present disclosure. It should be further noted that, for the convenience of description, only the portions relevant to the present disclosure are shown in the drawings.
It should be noted that the embodiments and features of the embodiments in the present disclosure may be combined with each other without conflict. The present disclosure will be described in detail below with reference to the accompanying drawings in conjunction with embodiments.
It should be noted that, the step numbers in the text are only for convenience of explanation of the specific embodiments, and do not serve to limit the execution sequence of the steps.
The methods provided by some embodiments of the present disclosure may be executed by a relevant processor, and are all described below by taking the processor as an example of an execution subject. The execution subject can be adjusted according to the specific case, such as a server, an electronic device, a computer, and the like.
CO2The mechanism of huff and puff development is complex, factors influencing the huff and puff effect are numerous, and in the related technology at present, the CO is about to be carried out2The oil field block developed by huff and puff has not constructed a complete set of CO2Throughput development effect evaluation method. The present disclosure is directed to horizontal well CO2The problem of evaluation of huff and puff development effect and creativity is solved by providing a CO establishment method based on artificial intelligence2Method of throughput effect prediction model and CO2A method for evaluating throughput effect.
In some embodiments of the disclosure, the oil change rate refers to the use of CO2Oil increase and CO injection for huff and puff development versus starve development2The mass ratio of (a). After the block development is finished, the oil change rate can be used as a representation of CO2Parameters of throughput effect. The larger the oil change rate is, the more CO is adopted in the well zone2The better the effect of the throughput mode.
As shown in FIG. 1, some embodiments of the present disclosure provide a method of establishing CO2The method of the throughput effect prediction model comprises S1-S5.
S1, obtaining CO2Taking the impact factor parameter data set and the target parameter data set developed in a throughput manner as sample data sets; wherein the influencing factor parameters comprise the soaking time, the fracture spacing, the water saturation, the formation pressure, the porosity, the permeability, the fracture length, the injection amount and the single-layer thickness, and the target parameters comprise the oil change rate.
The inventor finds that the CO is influenced in research2The factors of the throughput development effect mainly comprise the following aspects:
in terms of reservoir characteristics, influence of CO2Throughput developmentFactors of effect include fracture spacing, fracture length, monolayer thickness, and formation pressure. Wherein, at a certain oil layer temperature, the larger the formation pressure is, the CO is2The greater the solubility, the easier it is to enhance the mobility of the formation crude oil, while at the same time the swept volume can be enlarged.
In terms of rock properties, influence of CO2Factors that contribute to the throughput development effect include porosity, permeability, and oil saturation. Wherein, the oil change rate and the oil layer permeability level are in logarithmic relation, and the permeability of the ultra-low permeability reservoir (the permeability is less than 1 multiplied by 10)-3μm2) The permeability of (a) has a significant effect on the oil change rate; the larger the saturation of the remaining oil, the larger the oil gas yield, the lower the water content and the CO2The more advantageous the throughput.
Influencing CO in developing process parameters2Factors for throughput development effect include shot size and soak time. The larger the injection quantity is, the larger the huff and puff extraction degree is, but the amplification is gradually weakened, and the oil change rate is gradually reduced after being increased; the longer the soaking time, the greater the oil production, but the amplification gradually decreases.
For already performed CO2And in the well area developed by throughput, the data of the influencing factor parameters are relatively easy to obtain. For example, the soaking time, the fracture spacing, the water saturation, the formation pressure, the porosity, the permeability, the fracture length, the injection amount, the single-layer thickness and the oil change rate data of the related well zone can be directly derived in numerical simulation software, 10 parameters are totally derived, 516 groups of samples are derived from each parameter, the first 9 parameters are used as influencing factor parameters, and the oil change rate is used as a target parameter, so that the CO is obtained2The developed sample data set is throughput.
And S2, dividing the sample data set into a training set and a test set.
S3, building a neural network model, setting weights of neurons in the neural network model, and setting an activation function and an optimizer.
And S4, inputting the training set into the neural network model for training.
S5, if the loss function value of the neural network model is larger than the preset error, modifying the training times, the activation function, the optimizer and the Dropout ratio, and thenTraining the model; if the loss function value of the neural network model is within the preset error range, stopping training and taking the trained neural network model as CO2And (4) a throughput effect prediction model.
Some embodiments of the disclosure provide for establishing a CO2The method of the throughput effect prediction model comprises the steps of firstly obtaining CO2And handling a developed sample data set, dividing the sample data set into a training set and a testing set, setting a neural network structure, initializing hyper-parameters of a neural network model, training the model, and controlling a loss function value of the model within an error range by modifying the hyper-parameters, thereby obtaining the trained model. The neural network model trained by the method can be applied to treat CO under different influence factor parameters2Stimulation development oil change rate of a well zone is rapidly predicted, and compared with the result of numerical simulation prediction, CO is established through the method2The neural network model obtained by training the throughput effect prediction model is higher in prediction accuracy, stronger in adaptability and higher in calculation speed.
As shown in FIG. 2, in some embodiments, before step S2, the CO is established2The method of throughput effectiveness prediction model further includes S6.
And S6, processing the sample data set into a sample data set which can be used by machine learning, and normalizing the data in the processed sample data set.
For example, the sample data set can be converted into a Matlab file to become the sample data set which can be used by machine learning.
The data in the processed sample data set are normalized, so that the influence of the dimension on the final result can be eliminated, and the problem of gradient disappearance or gradient explosion of the neural network is prevented.
For example, the normalization process of the data in step S6 may use a Min-Max Scaling (Min-Max Scaling) method to convert the original data into a [0, 1] range.
The min-max normalization is:
Figure BDA0003224165180000061
wherein X is the original data, Xmin、XmaxMinimum and maximum values, X, respectively, of the original data setnomFor normalized data, such normalized values may be mapped to the interval [0, 1]]。
In some embodiments, in step S2, dividing the sample data set into a training set and a test set includes: the training and test sets are partitioned using the train _ test _ split function in the python open source library sklern.
Illustratively, the first 80% of the normalized sample data set is used as the training set and the last 20% is used as the test set, using the train _ test _ split function in the python open source library skleran.
The neural network structure is set to determine the number of layers of the neural network and the number of neurons per layer. In some embodiments, in step S3, the constructed neural network model includes an input layer, five hidden layers and an output layer, which are sequentially connected from the input end to the output end; wherein each hidden layer comprises 200 neurons. The neural network with the structure has higher prediction precision.
The feed-forward operation of the neural network of the architecture can be described as:
Figure BDA0003224165180000062
Figure BDA0003224165180000063
where f (-) represents the activation function, i is the index of the hidden layer of the network, and j is the index of the neuron of the hidden layer. z represents the input of the ith layer, y represents the output of the ith layer, and w and b represent the weight and offset of the ith layer, respectively.
In some embodiments, setting weights of neurons in the neural network model, selecting an activation function and an optimizer in step S3 includes:
and setting the weight of the neuron in the neural network model by adopting an Xavier method, adopting a ReLU function as an activation function, and selecting an Adam optimizer, thereby constructing a robust model.
The activation function adopts a ReLU function, can overcome the problem of gradient disappearance, and can accelerate the training speed, and the expression is as follows:
f(x)＝max(0,x)。
the weight initialization of the neural network model adopts an Xavier method, so that the activation values of all layers can not be saturated, and the activation values of all layers are not 0, thereby ensuring that the state variance and the gradient variance are kept unchanged.
Xavier initializes a weight variance as:
Figure BDA0003224165180000071
wherein n is the number of neurons, and the weight w of the ith layeriCan be initialized with a gaussian distribution as:
Figure BDA0003224165180000072
the Adam optimizer is adopted as follows:
calculating the gradient of t time step:
Figure BDA0003224165180000073
updating the biased first moment estimation:
mt←β1·mt-1+(1-β1)·gt；
and updating the estimation of the biased second-order original moment:
Figure BDA0003224165180000074
calculating the first moment estimation of deviation correction:
Figure BDA0003224165180000075
calculating the second-order original moment estimation of deviation correction:
Figure BDA0003224165180000076
updating parameters:
Figure BDA0003224165180000077
wherein t is a time step;
gtis the gradient at the time step t,
Figure BDA0003224165180000078
θt-1a parameter vector for a t-1 time step;
f (theta) is a random objective function of the parameter theta;
alpha is the step length and is 0.001 by default;
β1,β2e [0,1) meaning the exponential decay rate of the moment estimate, β1Default value is 0.9, beta2Default value is 0.999;
m is a first order moment vector;
v is a second moment vector;
ε is a parameter that prevents the denominator from being 0 and is typically set to 10-8。
In some embodiments, in step S5, the loss function of the neural network model includes Mean Squared Error (MSE). Further, step S5 includes: if the mean square error value of the neural network model is greater than 10-2Modifying the training times, the activation function, the optimizer and the Dropout ratio, and training the model again; if neural networkThe mean square error value of the model is less than or equal to 10-2Stopping training and using the trained neural network model as CO2And (4) a throughput effect prediction model.
The mean square error as a function of the loss is
Figure BDA0003224165180000081
Wherein, yiThe actual value is represented by the value of,
Figure BDA0003224165180000082
representing the predicted value, and n is the number of samples.
In the model training process, the activation function can be modified into a softmax function or a tanh function; the optimizer can be modified to Adadelta or SGD to control the loss function value of the model within the error range through multiple training, resulting in a trained model.
As shown in FIG. 2, in some embodiments, the CO is established2The method of the throughput effect prediction model further includes S7-S8.
S7, inputting the test set into CO2And predicting the oil change rate in the throughput effect prediction model.
S8, if the mean square error value of the predicted oil change rate is less than 10-2And determines the coefficient R2If the value is greater than 90%, the CO is considered2The throughput effect prediction model has higher prediction capability, namely CO2The prediction capability of the throughput effect prediction model meets the prediction requirement.
Some embodiments of the disclosure also provide a CO2The throughput effect evaluation method, as shown in fig. 3, includes S10 to S20.
S10, acquiring target well CO2And (3) throughput developing a data set of influencing factor parameters, wherein the influencing factor parameters comprise soaking time, fracture spacing, water saturation, formation pressure, porosity, permeability, fracture length, injection amount and single-layer thickness.
Here, the target well zone means that CO is to be subsequently conducted2Wells developed throughout.
S20, inputting the influencing factor parameter data set of the target well area into CO2A huff and puff effect prediction model for predicting the oil change rate; wherein, CO2The throughput effect prediction model is the establishment of the CO according to any of the embodiments2The model is established by a throughput effect prediction model method.
Some embodiments of the disclosure use CO2Taking the oil change rate of the huff-puff well zone as a prediction index, and establishing CO by the method2The neural network model built by the method of the huff and puff effect prediction model can quickly and effectively predict the CO carried out on the target well region under different influence factor parameters2Oil change rate of huff and puff development, so that CO in a target well zone is rapidly known2Expected effect of throughput development. The method has the advantages of high prediction accuracy, strong adaptability and high calculation speed. The process contributes to CO2And providing basis for handling well selection and adjustment of a later development scheme.
As shown in FIG. 4, in some embodiments, the CO2The throughput effectiveness evaluation method further includes S30.
And S30, analyzing the sensitivity of the oil change rate to different influencing factor parameters by using a Sobol sensitivity analysis method.
The influence weight of input parameters such as soaking time, water saturation, formation pressure, porosity, permeability, single-layer thickness parameters and the like on a prediction result can be determined by using a Sobol global sensitivity analysis method, so that influence factors or influence factor combinations with large influence on the oil change rate of a target well zone can be found, and influence on CO is facilitated2Analyzing the factors of the throughput capacity, thereby realizing the CO of the target well area2And (4) predicting and evaluating the throughput development effect. In addition, according to the analysis result, the subsequent development scheme of the target well zone can be adjusted by adjusting the corresponding parameter values of the influencing factors, and the analysis and guidance of the actual oil field development engineering are facilitated.
For the Sobol sensitivity analysis method, a k-dimensional unit body omega is definedk＝(x|0≤xiLess than or equal to 1; i 1,2, …, k), decomposing the function f (x) into 2pIncreasing the sum of terms to generateAnd (4) realizing data sampling by the Sobol random sequence, then calculating the total variance and each partial variance of the influencing factor parameters on the model response, and solving the sensitivity of each influencing factor parameter. When inputting parameter field IpFor p-element, the function f (x) is decomposed into 2pSum of the incremental terms:
Figure BDA0003224165180000091
in the formula (f)0As a constant, the integral of the other term over any of the included variables must be 0, i.e.:
Figure BDA0003224165180000092
all terms in the equation are orthogonal and can be expressed as the integral of the function f (x):
Figure BDA0003224165180000093
Figure BDA0003224165180000094
Figure BDA0003224165180000095
by analogy, other high-order terms in the formula (1) can be obtained. Squaring both sides of formula (1) and over the entire parameter domain IpInternal integration, in conjunction with equation (2), yields:
Figure BDA0003224165180000096
the total variance D of the function f (x) is
Figure BDA00032241651800001012
The variance is:
Figure BDA0003224165180000101
obtained by the formula (6):
Figure BDA0003224165180000102
the global sensitivity index is:
Figure BDA0003224165180000103
in the formula (I), the compound is shown in the specification,
Figure BDA0003224165180000104
is the ithkFirst order global sensitivity index of individual parameter for quantitative description of parameter
Figure BDA0003224165180000105
The influence of changes in (b) on the output results of the model;
Figure BDA0003224165180000106
is a second-order global sensitivity index for quantitatively describing parameters
Figure BDA0003224165180000107
The influence of the change on the output result of the model at the same time;
by analogy, defining total-order global sensitivity index si TThe sum of the sensitivity coefficients of each order of the parameters not only reflects the influence generated when the parameters are changed independently, but also reflects the influence generated when the parameters interact with other parameters. Dividing the set of factors x into x～iAnd xiThen si TComprises the following steps:
si T＝si+si(～i)＝1-s～i (11)
s～iis all that do not include the ith factor
Figure BDA0003224165180000108
Sum of terms, hence factor xiThe effect on the total variance of the output is:
si T＝1-D～i/D (12)
in the formula, D～iIs divided by a parameter xiOther parameters than these act together to produce a variance in the output of the model.
Will f is0，D，Di，
Figure BDA0003224165180000109
Estimated by monte carlo integration:
Figure BDA00032241651800001010
Figure BDA00032241651800001011
Figure BDA0003224165180000111
Figure BDA0003224165180000112
where n is the number of samples of the Monte Carlo estimate;
xmis omegakSampling points in space, x(～i)m＝(x1m,x2m,…,x(i-1)m,x(i+1)m,…,xkm)；
Superscript (1) (2) in equation (15) represents two n × k dimensional sample arrays of x.
CO provided by some embodiments of the present disclosure2The huff and puff effect evaluation method can be applied to rapid prediction of oil change rate under different influence factor parameters, has strong prediction accuracy and adaptability and high calculation speed, and can influence CO by a sensitivity analysis method based on a prediction result2Analyzing the factors of throughput to obtain CO2The evaluation of the effect of the throughput development has a guiding function.
Taking the oil field yellow 3 area of the Changqing highland as an example, the CO provided by some embodiments of the present disclosure will be described in detail2Method for establishing throughput effect prediction model and CO2A method for evaluating throughput effect.
The reservoir stratum of the yellow 3 region has poor physical property, strong heterogeneity and micro-crack development, but the crude oil has better property, low viscosity, low freezing point and stronger fluidity. Wherein the average buried depth of the long 8 oil layers is 2750m, and the thickness of the oil layer group is about 25-40 m. The reservoir is subdivided into 8 small layers, the main productive layer is 811 long, the average effective thickness is 13m, the porosity is 7.1%, the effective permeability is 0.39mD, and the reservoir belongs to a low-porosity and ultra-low permeability reservoir. Developing CO in the yellow 3 area as a pilot test area2Flooding and sequestration experiments, with CO starting in 20172The implantation of (2).
The oil deposit temperature of the well region is 84 ℃, the original formation pressure of the oil deposit is 25.9MPa, the saturation pressure is 8.48MPa, the initial oil saturation of the oil deposit is 60 percent, and the density of underground crude oil is 0.7248g/cm3The viscosity was 1.81 mPas. Formation water type CaCl2And the total mineralization is 35.42 g/L. CO implementation after horizontal well volume fracturing2Huff and puff development, the length of a horizontal well is 1200m, the half length of a crack is 160-140 m, and CO is injected periodically21500t, soaking for 30-60 days, and the whole operation on site is stable.
CO of the well2The parameters of the throughput influencing factors comprise the soaking time, the water saturation, the formation pressure, the porosity, the permeability, the fracture interval, the fracture length, the injection amount and the single-layer thickness. Based on the developed data, a sample data set corresponding to each influencing factor parameter is acquired through numerical simulation software, such as CMG or ECLIPSE, and 2592 groups of sample numbers are generated.
And carrying out Min-Max normalization processing on each parameter data in the sample data set.
And dividing the first 80% of the normalized sample data set into a training set and the last 20% of the normalized sample data set into a test set by using a train _ test _ split function in a python open source library skleern.
The method comprises the steps of setting a neural network structure, and setting the neural network as an input layer, an output layer and five hidden layers. The number of neurons per layer of the hidden layer was 200. And initializing the weight of the neural network, setting by adopting an Xavier method, selecting a ReLU activation function and adopting an Adam optimizer, and then inputting the normalized data into the built neural network model.
Training the model until the MSE value is less than 10-2And obtaining the built neural network model at the moment.
And predicting the oil change rate by using the model to predict the test set. The results show that the model predicts an oil change rate of 0.1432, an actual oil change rate of 0.1469, an MSE value of 0.0024 and R289.4 percent, so the neural network model has stronger prediction capability and can be used as CO2And (4) a throughput effect prediction model, and the prediction capability meets the prediction requirement.
And (3) analyzing the sensitivity degree of the oil change rate to different influencing factor parameters by using a Sobol sensitivity analysis method. Referring to fig. 5 and 6, DK represents monolayer thickness, FL represents fracture half-length, INJ represents injection amount, PERM represents permeability, POR represents porosity, PRES represents formation pressure, SW represents oil saturation, FS represents fracture spacing, and SOAK represents SOAK time.
As shown in FIG. 5, from the first-order sensitivity analysis results, the sensitivity of the oil change rate to the influencing factor parameters is ranked as: the soaking time is greater than the crack spacing, the oil saturation, the formation pressure, the porosity, the permeability, the half-length of the crack and the injection amount by a single-layer thickness.
As shown in FIG. 6, it can be seen from the full-order sensitivity analysis that the coupling effect of the monolayer thickness, the half length of the fracture, the injection amount, the oil saturation and other parameters is strong, the permeability, the porosity, the formation pressure, the fracture spacing and the soaking time are relatively independent, and the coupling degree with other parameters is small.
The analysis results adjusted CO for the subsequent yellow 3 region2And providing a basis for adjusting the priority order of the influencing factor parameters when the scheme is developed. Illustratively, if the predicted oil change rate in the yellow 3 zone does not reach the expected oil change rate, the soak time in subsequent development scenarios may be preferentially adjusted to bring the oil change rate to the expected value, thereby achieving better CO2Throughput development effects. For another example, when adjusting the injection amount in the subsequent development scheme, the coupling effect of the injection amount and other parameters is considered at the same time, and the oil change rate is made to reach the expected value by adjusting the values of a plurality of coupled parameters.
It is worth mentioning that the adjusted influencing factor parameters of the yellow 3 region can be input into the constructed CO2In the throughput oil change rate prediction model, the oil change rate of the adjusted development scheme is predicted through the model, so that a basis is provided for adjustment of the subsequent development scheme, and time and cost are saved. In addition, various influencing factor parameter adjustment schemes can be input into the CO2And predicting the oil change rate in the huff and puff oil change rate prediction model, and selecting a development scheme with the highest oil change rate or the most obvious effect of improving the oil change rate for subsequent development.
Some embodiments of the disclosure also provide a CO2The throughput performance evaluation apparatus 100 includes a processor 101 and a memory 102.
The processor 101 is used to support CO2The throughput effect evaluation apparatus 100 executes the CO according to any of the above-described embodiments2And a step in the throughput effect evaluation method. The processor 101 may be a Central Processing Unit (CPU), or may be other general-purpose processor, a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field Programmable Gate Array (FPGA) or other programmable logic device, a discrete gate or transistor logic device, a discrete hardware component, or the like. Wherein a general purpose processor may be a microprocessor or the processor may be any conventional processor or the like.
The memory 102 has stored therein computer program instructions adapted to be executed by the processor 101, the computer program instructions when executed by the processor 101 performing any of the aboveCO according to one embodiment2And a step in the throughput effect evaluation method.
The Memory 102 may be a Read-Only Memory (ROM) or other type of static storage device that can store static information and instructions, a Random Access Memory (RAM) or other type of dynamic storage device that can store information and instructions, an Electrically Erasable Programmable Read-Only Memory (EEPROM), a Compact Disc Read-Only Memory (CD-ROM) or other optical Disc storage, optical Disc storage (including Compact Disc, laser Disc, optical Disc, digital versatile Disc, blu-ray Disc, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium that can be used to carry or store desired program code in the form of instructions or data structures and that can be accessed by a computer, but is not limited to such. The memory 102 may be self-contained and coupled to the processor 101 via a communication bus. The memory 102 may also be integrated with the processor 101.
In the description herein, reference to the description of the terms "one embodiment/mode," "some embodiments/modes," "example," "specific example," or "some examples," etc., means that a particular feature, structure, material, or characteristic described in connection with the embodiment/mode or example is included in at least one embodiment/mode or example of the application. In this specification, the schematic representations of the terms used above are not necessarily intended to be the same embodiment/mode or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments/modes or examples. Furthermore, the various embodiments/aspects or examples and features of the various embodiments/aspects or examples described in this specification can be combined and combined by one skilled in the art without conflicting therewith.
Further, in the description of the present disclosure, "a plurality" means at least two, e.g., two, three, etc., unless explicitly specifically limited otherwise. Meanwhile, in the description of the present disclosure, unless otherwise explicitly specified or limited, the terms "connected" and "connected" should be interpreted broadly, e.g., as being fixedly connected, detachably connected, or integrally connected; the connection can be mechanical connection or electrical connection; may be directly connected or indirectly connected through an intermediate. The specific meaning of the above terms in the present disclosure can be understood by those of ordinary skill in the art as appropriate.
It will be understood by those skilled in the art that the foregoing embodiments are merely for clarity of illustration of the disclosure and are not intended to limit the scope of the disclosure. Other variations or modifications may occur to those skilled in the art, based on the foregoing disclosure, and are still within the scope of the present disclosure.

Claims (10)

1. Establishment of CO2A method for a throughput effectiveness prediction model, the method comprising:
obtaining CO2Taking the impact factor parameter data set and the target parameter data set developed in a throughput manner as sample data sets; the parameters of the influencing factors comprise the soaking time, the fracture interval, the water saturation, the formation pressure, the porosity, the permeability, the fracture length, the injection amount and the single-layer thickness, and the target parameters comprise the oil change rate;
dividing the sample data set into a training set and a test set;
building a neural network model, setting weights of neurons in the neural network model, and setting an activation function and an optimizer;
inputting the training set into the neural network model for training;
if the loss function value of the neural network model is larger than a preset error, modifying the training times, the activation function, the optimizer and the Dropout ratio, and training the model again; if the loss function value of the neural network model is within a preset error range, stopping training and taking the trained neural network model as CO2And (4) a throughput effect prediction model.
2. The establishment CO of claim 12The method for predicting throughput effect is characterized in that before the sample data set is divided into a training set and a test set, the CO is established2The method of the throughput effectiveness prediction model further comprises: and processing the sample data set into a sample data set which can be used by machine learning, and performing normalization processing on data in the processed sample data set.
3. The establishment CO of claim 12The method for predicting throughput effect is characterized in that the step of dividing the sample data set into a training set and a test set comprises the following steps: dividing a training set and a testing set by using a train _ test _ split function in a python open source library skleern; the first 80% of the sample data set is used as a training set, and the last 20% is used as a test set.
4. The establishment CO of claim 12The method for the throughput effect prediction model is characterized in that the built neural network model comprises an input layer, five hidden layers and an output layer which are sequentially connected from an input end to an output end; wherein each hidden layer comprises 200 neurons.
5. The establishment CO of claim 12The method for predicting the throughput effect is characterized in that the setting of the weight of the neuron in the neural network model and the selection of the activation function and the optimizer comprise the following steps:
and setting the weight of the neuron in the neural network model by adopting an Xavier method, adopting a ReLU function as an activation function, and selecting an Adam optimizer.
6. The establishment CO of claim 12A method of throughput effect prediction modeling, wherein a loss function of the neural network model comprises a mean square error;
if the mean square error value of the neural network model is greater than 10-2Then modify the training times and excitationLive functions, optimizers and Dropout ratios, and training the model again; if the mean square error value of the neural network model is less than or equal to 10-2Stopping training and using the trained neural network model as CO2And (4) a throughput effect prediction model.
7. The establishment CO of claim 12Method for throughput effect prediction model, characterized in that said establishing CO2The method of the throughput effectiveness prediction model further comprises:
inputting the test set into the CO2Predicting the oil change rate in the huff and puff effect prediction model;
if the mean square error value of the predicted oil change rate is less than 10-2And determines the coefficient R2A value greater than 90%, then the CO2The prediction capability of the throughput effect prediction model meets the prediction requirement.
8. CO (carbon monoxide)2A throughput effect evaluation method, characterized by comprising:
obtaining target well area CO2The method comprises the steps of taking out and developing an influence factor parameter data set, wherein the influence factor parameters comprise soaking time, fracture spacing, water saturation, formation pressure, porosity, permeability, fracture length, injection amount and single-layer thickness;
inputting the influencing factor parameter dataset of the target well region into CO2A huff and puff effect prediction model for predicting the oil change rate; the CO is2The method for predicting the throughput effect is to establish CO according to any one of claims 1 to 72The model is established by a throughput effect prediction model method.
9. CO according to claim 82The method for evaluating throughput effect is characterized in that the CO is2The throughput effect evaluation method further includes:
and (3) analyzing the sensitivity degree of the oil change rate to different influencing factor parameters by using a Sobol sensitivity analysis method.
10. CO (carbon monoxide)2Throughput effect evaluation device, characterized in that the device comprises a processor and a memory, in which computer program instructions adapted to be executed by the processor are stored, which computer program instructions, when executed by the processor, perform the CO of claim 8 or 92And a step in the throughput effect evaluation method.
CN202110966381.9A 2021-08-23 2021-08-23 Establishment of CO2Method of throughput effect prediction model and CO2Method for evaluating throughput effect Pending CN113592194A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202110966381.9A CN113592194A (en) 2021-08-23 2021-08-23 Establishment of CO2Method of throughput effect prediction model and CO2Method for evaluating throughput effect

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202110966381.9A CN113592194A (en) 2021-08-23 2021-08-23 Establishment of CO2Method of throughput effect prediction model and CO2Method for evaluating throughput effect

Publications (1)

Publication Number Publication Date
CN113592194A true CN113592194A (en) 2021-11-02

Family

ID=78239019

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202110966381.9A Pending CN113592194A (en) 2021-08-23 2021-08-23 Establishment of CO2Method of throughput effect prediction model and CO2Method for evaluating throughput effect

Country Status (1)

Country Link
CN (1) CN113592194A (en)

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114781951A (en) * 2022-06-23 2022-07-22 中国石油大学(华东) Shale oil reservoir carbon dioxide huff-puff development well selection decision method and system
CN115732041A (en) * 2022-12-07 2023-03-03 中国石油大学(北京) Carbon dioxide capture amount prediction model construction method, intelligent prediction method and device
CN115822533A (en) * 2022-11-28 2023-03-21 新疆敦华绿碳技术股份有限公司 Oil displacement method for bottom water sandstone reservoir
CN116066066A (en) * 2021-12-22 2023-05-05 中国石油天然气集团有限公司 Method for screening nitrogen throughput measure well
CN118155743A (en) * 2024-05-13 2024-06-07 中国石油大学(华东) CO based on image space moment theory2Plume distribution mode evaluation method

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108252688A (en) * 2018-01-08 2018-07-06 中国地质大学（北京） Compact oil reservoir CO_2 stimulation influence factor Sensitivity Analysis and its application
CN108960546A (en) * 2017-05-19 2018-12-07 中国石油化工股份有限公司 A kind of CO2 handles up well choosing method and device
CN111882033A (en) * 2020-07-15 2020-11-03 南京航空航天大学 Keras-based regional civil aviation active and passive carbon emission prediction method
CN112633328A (en) * 2020-12-04 2021-04-09 北京科技大学 Dense oil reservoir transformation effect evaluation method based on deep learning

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN108960546A (en) * 2017-05-19 2018-12-07 中国石油化工股份有限公司 A kind of CO2 handles up well choosing method and device
CN108252688A (en) * 2018-01-08 2018-07-06 中国地质大学（北京） Compact oil reservoir CO_2 stimulation influence factor Sensitivity Analysis and its application
CN111882033A (en) * 2020-07-15 2020-11-03 南京航空航天大学 Keras-based regional civil aviation active and passive carbon emission prediction method
CN112633328A (en) * 2020-12-04 2021-04-09 北京科技大学 Dense oil reservoir transformation effect evaluation method based on deep learning

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
侯天江等: "神经网络的CO2吞吐候选井模糊优选方法", 《西南石油学院学报》 *
张娟等: "水平井CO_2吞吐增油机理及影响因素", 《油田化学》 *
赵庶旭等: "《神经网络——理论、技术、方法及应用》", 31 May 2013, 中国铁道出版社 *

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116066066A (en) * 2021-12-22 2023-05-05 中国石油天然气集团有限公司 Method for screening nitrogen throughput measure well
CN116066066B (en) * 2021-12-22 2024-05-24 中国石油天然气集团有限公司 Method for screening nitrogen throughput measure well
CN114781951A (en) * 2022-06-23 2022-07-22 中国石油大学(华东) Shale oil reservoir carbon dioxide huff-puff development well selection decision method and system
CN115822533A (en) * 2022-11-28 2023-03-21 新疆敦华绿碳技术股份有限公司 Oil displacement method for bottom water sandstone reservoir
CN115822533B (en) * 2022-11-28 2024-09-20 新疆敦华绿碳技术股份有限公司 Oil displacement method for bottom water sandstone oil reservoir
CN115732041A (en) * 2022-12-07 2023-03-03 中国石油大学(北京) Carbon dioxide capture amount prediction model construction method, intelligent prediction method and device
CN115732041B (en) * 2022-12-07 2023-10-13 中国石油大学(北京) Carbon dioxide capture quantity prediction model construction method, intelligent prediction method and device
CN118155743A (en) * 2024-05-13 2024-06-07 中国石油大学(华东) CO based on image space moment theory2Plume distribution mode evaluation method

Similar Documents

Publication Publication Date Title
CN113592194A (en) Establishment of CO2Method of throughput effect prediction model and CO2Method for evaluating throughput effect
Erofeev et al. Prediction of porosity and permeability alteration based on machine learning algorithms
CN106351651B (en) Shale gas well productivity prediction method and device
Ahmadi et al. Robust intelligent tool for estimating dew point pressure in retrograded condensate gas reservoirs: application of particle swarm optimization
Zolotukhin et al. Machine learning in reservoir permeability prediction and modelling of fluid flow in porous media
CN116611350B (en) Oil well acidizing measure effect prediction method based on transfer learning
US12019202B2 (en) Fast variogram modeling driven by artificial intelligence
CN113792479A (en) Physical constraint-based coalbed methane reservoir fracturing effect evaluation method
CN114693005A (en) Three-dimensional underground oil reservoir dynamic prediction method based on convolution Fourier neural network
Raji et al. A new approach in well placement optimization using metaheuristic algorithms
Yaralidarani et al. An improved Ant Colony Optimization (ACO) technique for estimation of flow functions (kr and Pc) from core-flood experiments
Fan et al. Estimation of three-phase relative permeabilities for a water-alternating-gas process by use of an improved ensemble randomized maximum-likelihood algorithm
Muther et al. Subsurface physics inspired neural network to predict shale oil recovery under the influence of rock and fracture properties
Seyyedattar et al. Relative permeability modeling using extra trees, ANFIS, and hybrid LSSVM–CSA methods
Hatamizadeh et al. Simulation of carbonate reservoirs acidizing using machine and meta-learning methods and its optimization by the genetic algorithm
CN110717270A (en) Oil and gas reservoir simulation method based on data
US10934812B2 (en) Integrated a priori uncertainty parameter architecture in simulation model creation
Aryanto et al. Hydraulic fracturing candidate-well selection using artificial intelligence approach
Wang et al. On the Feasibility of An Ensemble Multi-Fidelity Neural Network for Fast Data Assimilation for Subsurface Flow in Porous Media
CN111751878A (en) Method and device for predicting transverse wave velocity
Fathi et al. Simultaneous well spacing and completion optimization using an automated machine learning approach. A case study of the Marcellus Shale reservoir, northeastern United States
Lieberman et al. Nonlinear goal-oriented Bayesian inference: application to carbon capture and storage
US20230184061A1 (en) Machine Learning with Physics-based Models to Predict Multilateral Well Performance
Sheng et al. Interpretable knowledge-guided framework for modeling reservoir water-sensitivity damage based on Light Gradient Boosting Machine using Bayesian optimization and hybrid feature mining
Gowida et al. Comparative analysis between different artificial based models for predicting static Poisson’s ratio of sandstone formations

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
RJ01 Rejection of invention patent application after publication

Application publication date: 20211102

RJ01 Rejection of invention patent application after publication